Interference Signal Recognition Based on Multi-Modal Deep Learning

Abstract

With the development of communication countermeasure technology, the future communication systems will face multi-source and multi-pattern interference deliberately released by the enemy, and the electromagnetic environment is becoming more complex. Effective communication anti-jamming technology needs to identify the interference signal of the enemy. Therefore, an interference signal recognition method based on multi-modal deep learning is proposed in this paper. Recently, multi-modal deep learning has become more and more important in machine learning. However, it has few applications in the field of communication. At present, most of the CNN-based signal recognition methods that have been studied are only based on the unimodal feature of the signal. We exploit a signal contour stellar images-I/Q waveform multimodal fusion (CWMF) method to achieve interference signal recognition based on AlexNet networks and deep complex-valued neural network. After feature extraction from multimodal data using AlexNet networks and deep complex-valued neural network, To merge multimodal features, we use the feature fusion method. In addition, we use decision-based fusion to merge multimodal models to obtain better distinguishable features and realize a better recognition performance.